Template Matching Device Geometrical Change Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing template-matching methods fail to accurately specify the matching position when input signals undergo geometrical changes, such as rotation or scale changes, leading to pseudo matching or failure in both two-dimensional image and one-dimensional audio signals.
Innovation Solution
A template-matching apparatus that calculates and compares multiple degrees of similarity between a template image and partial images extracted from an input image, using color component-based characteristic amounts and residual calculations to determine the matching position with high accuracy, even under geometrical changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional template matching algorithms (cross-correlation, sequential similarity detection) are used to compute signal strength for matching, then the matching process is simple and fast, but the matching position cannot be accurately specified when the input signal undergoes geometrical changes such as rotation or scale changes
Solution Approach 1:
The patent segments the matching computation into multiple independent components: a first computing unit that calculates a first degree of similarity based on characteristic amounts (e.g., color histograms), a second computing unit that calculates a second degree of similarity based on residual amounts (pixel-by-pixel differences), and a third computing unit that combines these to produce a third degree of similarity. This segmentation allows each component to handle different aspects of the matching problem, improving accuracy under geometrical transformations while maintaining computational efficiency through specialized processing for each similarity measure.
Solution Approach 2:
The patent changes the parameters used for matching by introducing multiple degrees of similarity computation. Instead of relying on a single matching criterion, the system computes a first degree of similarity using characteristic amounts that are invariant to certain transformations, a second degree of similarity using residual amounts for fine-grained comparison, and a third degree of similarity that integrates both. This multi-parameter approach enables accurate matching position specification even when the input signal undergoes rotation or scale changes.
2Reliability
If multiple degrees of similarity are computed and combined to improve matching accuracy under geometrical changes, then the matching position can be accurately specified, but the computation complexity and processing time increase
Solution Approach 1:
The patent performs preliminary computation by calculating the first degree of similarity based on characteristic amounts (such as color histograms or other transformation-invariant features) before computing the second degree of similarity based on residual amounts. This preliminary action using transformation-invariant features quickly eliminates obviously mismatched regions, allowing the more computationally intensive residual-based comparison to be focused only on promising candidate regions, thereby reducing overall computation time while maintaining high reliability under geometrical transformations.
Solution Approach 2:
The patent applies different quality measures to different aspects of the matching process: the first degree of similarity uses characteristic amounts that provide robustness to global transformations, while the second degree of similarity uses residual amounts that provide precision for local pixel-level comparison. The third computing unit combines these local quality measures appropriately, allowing the system to achieve high reliability under transformation by leveraging the strengths of each local quality measure in its appropriate context.
Data Source
AI summary
A template-matching apparatus includes a first calculating unit calculating a first characteristic amount from the image information of a template image, and extracting unit extracting a partial image, a second calculating unit calculating for image information of the partial image a second characteristic amount, a third calculating unit calculating a residual amount from the image information of the template image and the partial image, a first computing unit finding a first degree of similarity, a second computing unit that finds a second degree of similarity based on the residual amount, a third computing unit finding a third degree of similarity based on the first and second degree of similarity, and a specifying unit specifying a matching position, thereby specifying the matching position with good accuracy even if the input image is observed with some geometrical change.